Boundaryless Visible Light Communication System Without Attocell for Industrial Internet of Things

Recently, visible light communication (VLC) in the industrial Internet of Things has been widely concerned. In these systems, multi-access technologies are always used to eliminate interference between adjacent transmitters, significantly decreasing transmission efficiency. Meanwhile, the attocell boundaries caused by multi-access technologies can also limit communication coverage and lead to frequent receiving switches. To overcome these weaknesses caused by the division of attocell, we design a boundaryless none-attocell VLC system in this paper. In the proposed system, each transmitter uses all available communication resources, and a cubic receiver is adopted to receive in both non-overlapping and overlapping areas. Specifically, we propose a scheme for adaptive signal selection, enabling the receiver to select the recoverable transmitting signals. After that, we introduce an adaptive channel estimation and signal recovery method, which adaptively achieves signal combining and overlapping signal recovery. Our simulations verify the proposed scheme and demonstrate that the proposed system with the adaptive receiving scheme has better error performance and broader communication coverage than the traditional attocell system.


I. INTRODUCTION
R ECENTLY, the utilization of visible light communication (VLC) in the Industrial Internet of Things (IIoT) has been widely studied because of VLC 's superior qualities, including endogenous security and high-efficiency [1]- [4]. In IIoTs, VLC transmitters are densely arranged to broadcast the status data of their covering area. The coverages of these transmitters always overlap with each other, leading to severe signal interference [5], [6]. Therefore, the concept of attocell has been introduced to overcome the interference [7], [8].
In attocell systems, multi-access technologies including frequency division multi-access (FDMA), time division multi-access, code division multi-access, etc. are always used between adjacent transmitters [9]- [14]. These technologies divide communication resources into orthogonal ones for different transmitters, and thus the communication efficiency is significantly decreased.
Meanwhile, the communication coverage is divided into several attocells without overlapping at the receiving end. However, these attocells' boundaries limit the communication coverage because the receiver could only receive from one transmitter at a time. Additionally, when the receiver moves to another attocell, handover is operated to match the new transmitter, which includes a series of control and feedback above the physical layer and could increase communication latency [15], [16].
To improve the communication efficiency and expand transmitters' coverage, we consider establishing a boundaryless VLC system removing the concept of attocell, where transmission is carried out without multi-access technologies, and adjacent transmitters' communication coverage can overlap. In this system, every transmitter transmits signals using all available communication resources without considering interference. The receiver can adaptively achieve antinoise receiving in non-overlapping areas and simultaneous recovery of two transmitting signals in overlapping areas. Additionally, handover is not required due to the cancellation of boundaries.
Consequently, the contributions of our paper are summarized as follows.
• First, we propose a boundaryless none-attocell system for IIoTs. In this system, a cubic receiver is designed for spatial channel estimation and diversity receiving from different directions [17]. We divide the signal receiving into non-overlapping and overlapping cases according to the receiver's location and then conclude both receiving cases into multi-input multi-output transmission problems. • We propose a receiving scheme adaptive to two cases, including signal selection and recovery. For signal selection, we design a multi-subframe format and offer a method for recoverable signal recognition. For adaptive signal recovery, we apply a spatial channel estimation based on the average receiving power of the cubic receiver, which adaptively estimates the channel matrix in both cases without receiving mode switch and channel training sequence. After that, an adaptive signal recovery method is also introduced. • Finally, we simulate the communication performance of the none-attocell transmission and adaptive receiving scheme. The results demonstrate that, because of the full-resource transmission and adaptive receiving, the boundaryless none-attocell transmission has a lower error rate and broader communication coverage than the traditional attocell system with flat receivers. The remainder of our paper is organized as follows. First, we introduce our system model in Section II. Then, we present an adaptive signal selection and recovery scheme in Section III. After that, some simulations are conducted in Section IV. Finally, we conclude the paper in Section V.

II. BOUNDARYLESS NONE-ATTOCELL SYSTEM MODEL
In this paper, we propose a boundaryless none-attocell VLC system as Fig. 1a. At the transmitting end, Lambertian transmitters are arranged on a straight line with the same interval D [18], [19]. These transmitters simultaneously broadcast signals modulated by different data using all available communication resources rather than multi-access technologies. Note that D is reasonably set to guarantee the communication and illumination coverage but not make the transmitters too intensive. A mobile user with a VLC receiver moves in transmitters' coverage at the receiving end, including non-overlapping area (Case A) and overlapping area (Case B). The receiver requires anti-noise receiving from the corresponding transmitter in Case A. In Case B, the user needs simultaneous recovery of two transmitting signals from the overlapping receiving signals for boundary cancellation. Note that to simplify the receiving process, we do not consider cases where coverage of more than two transmitters overlaps with each other. For Lambertian light sources, the light intensity far from the center of the light spot is significantly attenuated, making the above situation rarely happens. Even though there is a need for dense distribution of light sources, the mentioned situations can be avoided by adjusting the half-power angle of transmitters.  In order to achieve anti-noise receiving and multiple signal recovery, we adopt a CR equipped with five photon detectors (PDs) R i (i = 1, 2, . . . , 5) on its top and side surfaces as Fig.  1b to receive from different directions. CR's two opposite receiving surfaces have opposite normal vectors and can be used to simultaneously receive from two transmitters. The top receiving surface can receive from the nearest transmitter like the traditional flat receiver. The unique receiver structure is conducive to the completion of multi-input channel estimation and signal reception. The receiver is not rotary in moving process, keeping the normal vectors n 2 and n 4 parallel to centerline and n 5 vertical to floor.
According to CR 's multiple receiving feature, we analyze and design the system under MIMO model [20], [21]. Under ideal signal sampling, the transmission follows Y = C s HP t + N 0 , where Y is the vector of signal amplitudes received by five PD arrays, Cs is PD 's responsivity [22] , H is channel matrix without considering reflected light, P t is the vector of transmitting power by two transmitters, and N 0 is a zero-mean Gaussian noise vector. For ease of presentation, we convert the equation to where P r is the vector of receiving power from five PDs, and N is another zero-mean Gaussian noise vector. Next, we will introduce an adaptive signal selection and recovery scheme cooperated with CR for boundaryless none-attocell model.

III. ADAPTIVE SIGNAL SELECTION & RECOVERY SCHEME
The scheme of adaptive signal selection and recovery aims to achieve two functions. In signal selection, the receiver can adaptively select the transmitting signals to receive after the recoverable signals are recognized. In signal recovery, CR adaptively achieves anti-noise recovery of one transmitting signal in Case A and simultaneous recovery of two transmitting signals from overlapping signals in Case B.

A. ADAPTIVE SIGNAL SELECTION
In this part, we first design a frame format for adaptive signal selection, which enables the overlapped transmitting signals to be separable rather than multi-access technologies. Based on the frame format, we realize recoverable signal recognition by analysing signal components and CR positioning to serve the user's adaptive receiving selection.

1) Frame format design
To adapt the channel change in the user's movement, we propose a frame format as Fig. 2. This format consists of multiple subframes of (L l + L s ) bits, which is the unit of our signal processing. A constant sync header is configured at the head of the frame to detect the start of the transmitting signal. There is a transmitter label (TL) of L l bits at the end of each subframe, which helps to recognize the transmitters and estimate the channel. We construct all transmitter labels into a set L = {l T }, where T is the serial number of transmitters, noting that L is stored in each receiver. The correlation between different TLs is tiny. Additionally, an end-of-frame mark is also arranged.

2) Recoverable signal recognition for user 's adaptive receiving selection
Our main idea is that we first roughly position the user by figuring out its two adjacent transmitters T l and T r based on TL and then analyze the recoverable transmitting signal according to T l , T r and synchronization result. Hence, the user knows the recoverable transmitting signals and adaptively selects them according to their real-time location. Before we introduce the method, we define a correlation operation which is the basis of transmitter identification. Definition 1. There is a correlation operation where s i is the sample vector of a subframe length received by R i , j is the position of TL 's first bit in sampling sequence which is obtained by synchronization, andmod(·, L s ) denotes modular L s arithmetic. Now, we introduce the joint analysis of recoverable signal and user location method in three cases, noting that we operate synchronization by sliding correlation [23]. Since each subframe has a TL, CR can timely detect the change of transmitters and change the real-time signal selection scheme as long as a subframe is recently received, which is suitable for mobile VLC.

B. ADAPTIVE SIGNAL RECOVERY
In this part, we first propose a channel estimation method adaptive to both cases. Based on the estimated H, we offer a signal recovery method that adaptively realizes signal combining in Case A and overlapping signal recovery in Case B.

1) Adaptive channel estimation
To improve the transmitting efficiency and the adaptability to users' movement, this channel estimation method uses the average power of CR's multiple receiving signals in one subframe rather than the channel training sequence [24].
The method is designed according to Lambertian radiation feature and CR 's geometrical relationship. Here, we present the geometrical diagram of Case A and Case B as Fig. 3. We calculate the mean power of receiving signals in a subframe length and denote them asP r = P 1P2P3P4P5 T , whereP i corresponds to R i .
Then, we can estimate channel matrix of Case B using Theorem 1.
whereĥ 1 andĥ 2 are the estimated channel factors between T r , T l and CR. Set X 1 =ĥ 1 d 1 and X 2 =ĥ 2 d 2 , and then , and P 0 is the mean power of transmitting signals. Choose the feasible solution from two roots of X 1 and X 2 which satisfies [R (s 5 , l T l ) − R (s 5 , l Tr )] (X 1 − X 2 ) ≥ 0. Note that if T l is not synchronized with CR, we assume that R (s 5 , l T l ) = 0, and the same assumption goes for T r . Then, where A r is the detective area of PD array, and m 0 = −ln 2/ln ψ h (ψ h is the half-power semi-angle of transmitters). Then, other varieties are obtained as (6).
Meanwhile, Theorem 1 obeys Property 1, the proof of which is in Appendix B.
Property 1. When transmitter interval D is large enough, the channel estimation result for Case B is equal to Case A (Fig. 3a).
Therefore, our channel estimation method is adaptively suitable for any location of CR.

2) Adaptive signal recovery method
According to the estimated channel matrix, we could achieve anti-noise and anti-interference signal recovery based on the signal combination and minimum mean square error criterion [25].
In this algorithm, we first construct all the possible combinations of two transmitters' emitting power into a set X = {X i }. After that, we determine the transmitted power combinationX = min Then, we recover the transmitted data fromX according to the preset mapping relationship. For Case A, one of the data streams can be robustly recovered because of signal combining, and the other data stream is randomly determined, which would not be selected to receive due to recoverable signal analysis. For Case B, data from two adjacent transmitters are recovered simultaneously. In this way, signals are adaptively and boundaryless recovered in two cases.

IV. SIMULATION
In this section, we operate simulations to verify the communication enhancement including error rate and coverage of the proposed boundaryless none-attocell system and its adaptive receiving scheme. All of these simulations are operated in the scenario as Fig. 1a. The parameters of Simulation A and B are in Tab. 1 1 .

A. COMPARISON OF ADAPTIVE MULTI-CHANNEL RECEIVING & SINGLE CHANNEL RECEIVING
We compare the bit error rate (BER) when using CR-based adaptive multi-channel receiving and single-channel receiving, noting that single channel receiving uses a flat receiver with 5 PDs facing the ceiling. Here, T 1 ∼ T 4 transmit different signals using 2-PAM without multi-access technology under different R b . Because of the symmetry of the system, we only present BER of receiving from T 3 when x ∈ [−2, 2] in Fig. 4. As we can see, when the receiver is closer to T 3 , the BER of both methods is very low. However, when the receiver is closer to T 2 , the BER of SCR severely deteriorates, while AMCR performs far better to enable the receiving in most areas. The result illustrates that the CR-based adaptive multi-channel receiving method has broader communication coverage and can simultaneously receive from two transmitters.

B. COMPARISON OF NONE-ATTOCELL SYSTEM & ATTOCELL SYSTEM
Now we compare the BER distribution in the boundaryless none-attocell system with AMCR and the attocell system with SCR, which uses FDMA between adjacent transmitters. In the attocell system, every transmitter only uses half of the 1 PD 's parameters are referenced as HAMAMATSU S12915-1010R. available bandwidth, and adjacent transmitters use different frequency bands to cancel interference [26], [27]. We operate two sets of simulations. In each set, the total available bandwidths of two systems are the same, and two systems use different PAM orders to make R b = 10Mbps. As shown in Fig. 5, BER does not steeply change at x = 0 for both systems, indicating that both multi-access technology and none-attocell design expand the communication coverage. However, in both simulation sets, the none-attocell system has a lower error rate at the same receiver's location compared with the attocell system which uses frequency multiplexing technology. Although for the 4-PAM noneattocell system, BER fluctuates around x = 0 because of the wrong choice of feasible solutions for X 1 and X 2 , the effect is not significant since their difference is slight. Therefore, the proposed system could achieve broader communication coverage for each transmitter due to the full utilization of frequency bandwidth.

C. COVERAGE IMPROVEMENT OF BOUNDARYLESS NONE-ATTOCELL SYSTEM
Next, we compare the communication coverage of the noneattocell system and attocell system. Here, we adjust the value VOLUME 4, 2016 (c) For the simulated ψ h , it is obvious that the proposed scheme has broader communication coverage than the attocell system with FDMA. Meanwhile, as the interference increases, i.e. ψ h becomes larger, the coverage enhancement improves. This is because the proposed scheme can adaptively recover signals in the same frequency band even if they are overlapped, while the attocell system cannot. Consequently, our non-attocell system achieves broader communication coverage for each transmitter under interference, making it more adaptable to mobile users and providing more possibilities for the light source arrangement in IIoTs.

V. CONCLUSION
In this paper, we designed a boundaryless none-attocell VLC system for IIoT. In this system, every transmitter transmits without multi-access technology. A CR with adaptive receiving scheme boundaryless receives from one transmitter in non-overlapping areas or two transmitters in overlapping areas. Our simulations verified the BER decrease of adaptive boundaryless receiving and none-attocell transmission and demonstrated that our system has broader communication coverage. .

APPENDIX A PROOF OF THEOREM 1
In Fig. 3b, the geometrical relationship in (8) where D and H are known. Taking the expectation of (1), then EP r = EP 1 EP 2 EP 3 EP 4 EP 5 T = H P 0 1 T (9) where EP r is the expectation of P r . Here, we substitute EP i byP i . Then, (4) can be acquired with (8), (9) and (3). After that, because X 1 , X 2 and R (s 5 , l T ) are both negatively correlated with d, so [R (s 5 , l T l ) − R (s 5 , l Tr )] (X 1 − X 2 ) ≥ 0. Thus, the feasible X 1 and X 2 are obtained. Moreover, (5) is based on Lambertian model, and (6) is then acquired.